Update training_args.py (#12328)

mention in `save_strategy` param description that `load_best_model_at_end` can override

src/transformers/training_args.py

@@ -171,12 +171,12 @@ class TrainingArguments:
         logging_steps (:obj:`int`, `optional`, defaults to 500):
             Number of update steps between two logs if :obj:`logging_strategy="steps"`.
         save_strategy (:obj:`str` or :class:`~transformers.trainer_utils.IntervalStrategy`, `optional`, defaults to :obj:`"steps"`):
-            The checkpoint save strategy to adopt during training. Possible values are:
+            The checkpoint save strategy to adopt during training (Note that when :obj:`load_best_model_at_end=True`,
+            this parameter is ignored and the model is saved after each evaluation). Possible values are:
 
                 * :obj:`"no"`: No save is done during training.
                 * :obj:`"epoch"`: Save is done at the end of each epoch.
                 * :obj:`"steps"`: Save is done every :obj:`save_steps`.
-
         save_steps (:obj:`int`, `optional`, defaults to 500):
             Number of updates steps before two checkpoint saves if :obj:`save_strategy="steps"`.
         save_total_limit (:obj:`int`, `optional`):